This paper presents an empirical study comparing the effectiveness of four different versions of an on-line database tutorial, each of which calls upon the student to perform a different kind of inference. The general-to-specific version presents instructions in the form of general rules, from which the student s expected to infer how to apply the rule in the give context. The explanation-to-specific version supplies information about the functional organization of the database program in addition to general rules. The specific-to&-specific condition gives an example of the use of a command; the student must infer how to apply the command in a slightly different context. The control version gives explicit instructions. The best performance on a post-test consisting of realistic tasks was obtained from the general-to-specific and explanation-to-specific conditions.